If you’re searching for a place to live, one of the most important factors to consider is the time you’ll spend travelling between home and key locations, such as work or university.
With journey time being a critical factor when looking for a home, having the right tool to find the most suitable property is key. Luckily, companies like Alpaca are helping to make the search process easier.
Alpaca is a real estate technology company that enables renters to find their perfect home online. It owns the largest global network of housing renting groups worldwide, with over 10 million global members in its Facebook housing communities. Within its 1,000+ Facebook groups, Alpaca smartly connects renters with real estate agents and property managers through an AI-driven chatbot.
Since integrating the TravelTime API into its search recommendation algorithm, Alpaca’s chatbot can display the most relevant property listings, based on their proximity to specified points of interest.
Creating a personalised search user experience
Alpaca’s online communities span the globe, from the US to the UK, Germany, Italy and Australia. To effectively match renters with the most suitable properties, the team at Alpaca realised early on that journey time would be an important factor to include in their recommendation algorithm.
“In almost every user interview we conducted, people would say that they needed to know how long it would take them to commute from home to their workplace or university,” says Nicolas Beuchat, Co-Founder. “So, we needed to find a way to allow our users to easily see this information, as this often determines whether or not a viewing is requested.”
However, offering a search experience that is personalised and tailored to each user was a key challenge. “People may have very complex requirements and it’s hard to capture the full flexibility in an easy-to-use interface,” says Nicolas.
For example, a user may only want to live within 30 minutes of a point of interest, such as their workplace. But they may also want to be close to another point of interest, such as a school.
“Just for points of interest alone, we already have a lot of different needs and desires depending on the person,” says Nicolas. “If you then consider requirements about the property itself, that’s yet another layer of complexity. Creating a UI and recommendation algorithm that is generalised for everyone when there are criteria that are very specific to a few users is certainly a challenge.”
Enriching Alpaca’s search recommendation algorithm with travel time data
Integrated into Facebook Messenger is Alpaca’s AI-driven chatbot, Alex. The bot filters through thousands of listings to find those that are the most suited to the user’s requirements.
The chatbot’s recommendation algorithm uses the TravelTime API to determine how well connected a property is to a specified point of interest (POI) – and this is given a POI score. The higher the score, the higher the relevant listing appears in the app’s search results.
Thanks to the TravelTime API, the chatbot also displays journey times, in minutes, for each property listing based on the transport mode chosen. This includes walking, public transport, cycling and driving.
“A key benefit of using travel time data within our algorithm is that our users don’t have to use a separate website or app to check journey times. They can get this information directly from our chatbot,” says Sebastian Illing, Co-Founder.
Once a user engages, the chatbot sends daily recommendations, personalised to their preferences. When the user finds the right listing, they can use the chatbot to schedule remote or in-person viewings, at no extra fee.
How it works
Let’s take a deeper look at how Alpaca and the TravelTime API can simplify the house-hunting process.
When we first engage with the chatbot, we can choose to set our search preferences:
This opens a view in which we can enter important information, such as our preferred location, move-in date, and price.
Using the TravelTime API, the chatbot also allows us to enter one or multiple points of interest (POI) to see how far, in minutes, each property listing is from those POIs. Next, we can specify our commute preferences, with multiple transport options to choose from — driving, cycling, walking and public transport:
We can also specify other criteria, such as the number of bedrooms, a preference for furnished or unfurnished properties, as well as advanced criteria like air conditioning or laundry in unit. Once we’ve saved our preferences, we can enter our contact details to be kept up to date through personalised recommendations.
A call is made to the TravelTime API to get the travel times for the specified points of interest. Based on these travel times, Alpaca’s recommendation algorithm sorts and ranks the suitable property listings in order of relevance:
To visualise how far away a potential property is from our desired POI, we can also view our points of interest on a map:
Additionally, we can view a map with all the available listings and journey times in one place:
Determining the most relevant listings using travel time data
To help decide which listings to display, Alpaca’s recommendation algorithm uses a POI score, based on travel time data from the TravelTime API. Three different time-based criteria are used to calculate the POI score:
- Under 30 minutes: classified as being close to a point of interest;
- 45 minutes: classified as average;
- Over an hour: classified as too far, gets a negative score from the ranking algorithm.
As Nicolas explains, “In a city like New York, which is so competitive and expensive, we still want to show properties that may be a little outside of our users’ search criteria but are still relevant. This gives our users more relevant options that they may not have otherwise considered.”
Adding a human touch to the user experience is Alpaca’s Customer Success team. For example, if you are unfamiliar with a city, the team can help you identify whether you need to increase your budget or suggest other neighbourhoods that may be more accessible.
“We find that having a chatbot helps to automate the majority of our processes, but having a human touch is great for those more complex cases,” says Sebastian.
Journey time is a critical factor to ranking relevant search results
Apart from journey time, Alpaca’s recommendation algorithm weights additional criteria to help determine which listings to display. This includes the listings that people are actively interested in, with intent signals such as whether a viewing has been requested or whether a listing has been added to a user’s favourites.
The algorithm works from all the scores of these factors – for example, weighting how well the proposed amenities match with the commute time and rent price compared to other listings. Based on this weighting, regressions are run to find the right balance of weights for each.
“The POI score is based on data from the TravelTime API and it has a relatively high weight compared to other scores. This means that it’s a very important factor for our users,” says Sebastian. “We’re still experimenting to find the right balance. But we know that travel time is a critical factor.”
Increased user engagement and ease of use
Since implementing the TravelTime API, Alpaca has seen positive results. After running A/B tests on their recommendation algorithm, the team has found increased engagement with the point of interest functionality.
“We’ve found that our users really like the POI functionality. Whereas before we thought that factors like amenities would be most important for our users, we’ve found that the POI has a much higher usage from our users. Over 40% of our users currently use it,” says Sebastian.
Ease of use has been another benefit for the team. “The API itself is simple to use and really straightforward to implement,” says Sebastian. “We’ve found it to be very user-friendly and developer-friendly.”
Looking ahead, Alpaca has ambitious plans to expand to the rest of the US and other geographies, as well as other messenger platforms like WhatsApp and iMessage.
“Building on our work in personalised recommendations is a key priority for us,” says Sebastian. “Our approach has always been to offer a ‘passive’ search experience, where users don’t have to actively search for listings every day – we provide daily recommendations.
“This means that you have to trust that we can provide you with great recommendations and that you don’t need to scan all the listings yourself, since all of the listings we send fit your requirements. The first step is to nail down our users’ preferences. And then the next step is translating those into good recommendations. The TravelTime API will be a key part of this ongoing work.”
Create travel time polygons and matrices with the TravelTime API